[转][黄忠成]Object Builder Application Block (3)

[转]Object Builder Application Block

文/黄忠成 ；2006/9/21

原文链接：http://blog.csdn.net/code6421/archive/2006/09/25/1282161.aspx

整理：吕震宇

五、Misc

5-1、SingletonStrategy

SingletonStrategy可于对象实体首次建立后，将实体保留在Context中的Locator内的ILifetimeContainer对象中，之后相同型态、id相同的对象建立动作，都是传回这个对象，这是Singleton模式的实现，如程序27。

程序27

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_SingletonTest
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new SingletonStrategy());
context.InnerChain.Add(new CreationStrategy());
context.Policies.Set<ISingletonPolicy>(new SingletonPolicy(true), typeof(TestObject), null);
context.Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());

TestObject obj1 = (TestObject)context.HeadOfChain.BuildUp(context,
typeof(TestObject), null, null);
TestObject obj2 = (TestObject)context.HeadOfChain.BuildUp(context,
typeof(TestObject), null, null);
if (obj1 == obj2)
Console.WriteLine("Singleton");
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public class TestObject
{
}
}

要将一个『类型/id』标示为Singleton，设计者必须于Strategy串行中加入SingletonStrategy对象，并建立一个SingletonPolicy对象，这是一个实现了ISingletonPolicy接口的类别，其构造函数如下。

public SingletonPolicy(bool isSingleton);

CreatationStrategy在建立对象后，会从context.Policies中取出『类型/id』对应的ISingletonPolicy对象，以其IsSingleton属性来决定建立的对象是否为Singleton模式，是的话就将该对象实体填入ILifetimeContainer中，同时以DependencyResolutionLocatorKey包装该对象实体，放入Locator中，如下所示。

private void RegisterObject(IBuilderContext context, Type typeToBuild, object existing, string idToBuild)
{
if (context.Locator != null)
{
ILifetimeContainer lifetime = context.Locator.Get<ILifetimeContainer>(
typeof(ILifetimeContainer), SearchMode.Local);

if (lifetime != ;null)
{
ISingletonPolicy singletonPolicy = context.Policies.Get<ISingletonPolicy>(
typeToBuild, idToBuild);

if (singletonPolicy != null && singletonPolicy.IsSingleton)
{
context.Locator.Add(new DependencyResolutionLocatorKey(
typeToBuild, idToBuild), existing);
lifetime.Add(existing);
//...................
}
}
}
}

以上流程是当该对象实体尚未建立时的流程，假如以BuildUp建立的对象已经存在于Locator中，那么SingletonStrategy的BuildUp方法将直接传回Locator中的对象实体。

public override object BuildUp(IBuilderContext context, Type typeToBuild, object existing, string idToBuild)
{
DependencyResolutionLocatorKey key = new DependencyResolutionLocatorKey(
typeToBuild, idToBuild);

if (context.Locator != null && context.Locator.Contains(key, SearchMode.Local))
{
TraceBuildUp(context, typeToBuild, idToBuild, "");
return context.Locator.Get(key);
}
return base.BuildUp(context, typeToBuild, existing, idToBuild);
}

PS：注意，SingletonStrategy在该对象已经存在于Locator中时，是直接回传，并不会调用后面如MethodExecutionStrategy、PropertySetterStrategy等Strategy。

5-2、TypeMappingStrategy

前面的章节早已使用过TypeMappingStrategy这个对象了，它主要负责『类型/id』的对应，例如将IDataProcessor接口类型的建立，替换成PromptDataProcessor类型，如程序28所示。

程序28

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_TypeMappingTest
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new TypeMappingStrategy());
context.InnerChain.Add(new CreationStrategy());
ITypeMappingPolicy policy = new TypeMappingPolicy(typeof(TestObject), null);
context.Policies.Set<ITypeMappingPolicy>(policy, typeof(ITestInterface), null);
context.Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());
ITestInterface obj1 = (ITestInterface)context.HeadOfChain.BuildUp(
context, typeof(ITestInterface), null, null);
obj1.SayHello();
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public interface ITestInterface
{
void SayHello();
}

public class TestObject : ITestInterface
{
public void SayHello()
{
Console.WriteLine("TEST");
}
}
}

TypeMappingStrategy必须搭配TypeMappingPolicy对象使用，TypeMappingPolicy是一个实现ITypeMappingPolicy接口的对象，构造函数声明如下。

public TypeMappingPolicy(Type type, string id)

第一个参数是映像的实体类型，以本例来说就是TestObject，第二个参数是识别id，接着将其加入context.Policies中，如下所示。

context.Policies.Set<ITypeMappingPolicy>(policy, typeof(ITestInterface), null)

当TypeMappingStrategy的BuildUp方法被调用时，它会以『类型/id』取得对应的ITypeMappingPolicy对象，透过它来取得对应的类型，之后将使用这个类型调用下一个Strategy的BuildUp方法，这就是Type Mapping的流程。

PS:注意，Type Mapping类型必须兼容，如接口->实现、基础类别->衍生类别。

5-3、BuildAwareStrategy

BuildAwareStrategy可以于实现IBuilderAware接口对象建立或释放时，调用对应的OnBuildUp或OnTearDown方法，如程序29所示。

程序29

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_BuildAwareTest
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new CreationStrategy());
context.InnerChain.Add(new BuilderAwareStrategy());

context.Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());

TestObject obj = (TestObject)context.HeadOfChain.BuildUp(context,
typeof(TestObject), null, null);
context.HeadOfChain.TearDown(context, obj);
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public class TestObject : IBuilderAware
{
#region IBuilderAware Members

public void OnBuiltUp(string id)
{
Console.WriteLine("Object is build up");
}

public void OnTearingDown()
{
Console.WriteLine("Object is TearDown");
}

#endregion
}
}

与其它的Strategy对象不同，BuilderAwareStrategy并不需要Policy对象的协助，它只是判断建立的对象是否实现了IBuilderAware接口。

5-4、BuildUp的第三、四个参数

截至目前为止，我们的例子在调用BuildUp方法时，第三及四个参数都传入null，这两个参数的用途究竟为何呢？这要先从BuildUp方法的宣告谈起。

object BuildUp(IBuilderContext context, Type typeToBuild, object existing, string idToBuild);

当我们于调用BuildUp方法指定existing为一对象实体时，CreationStrategy将不会建立任何新的对象，只会进行Singleton模式对象的相关动作，然后就调用下一个Strategy对象的BuildUp方法，简单的说！在CreationStrategy后的Strategy仍然会运行，例如Method Injection、Setter Injection都会再次运行，程序30可以协助读者理解这两个参数的用途。

程序30

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_ExistingTest
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new CreationStrategy());

ConstructorPolicy policy = new ConstructorPolicy(new ValueParameter(typeof(string), "id"));
context.Policies.Set<ICreationPolicy>(policy, typeof(TestObject), null);

TestObject obj = (TestObject)context.HeadOfChain.BuildUp(context,
typeof(TestObject), null, null);
TestObject obj2 = (TestObject)context.HeadOfChain.BuildUp(context,
typeof(TestObject), obj, null);

if (obj == obj2)
Console.WriteLine("is same object.");
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public class TestObject
{
private string _id;

public string ID
{
get
{
return _id;
}
}

public TestObject(string id)
{
_id = id;
}
}
}

BuildUp的第四个参数则主导着ObjectBuilder的类型识别及对象识别机制，请先看程序31的例子。

程序31

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_IDTesting
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new CreationStrategy());
context.InnerChain.Add(new PropertySetterStrategy());

context.Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());

PropertySetterInfo pi1 = new PropertySetterInfo("ID", new ValueParameter<string>("ID1"));
PropertySetterPolicy pp1 = new PropertySetterPolicy();
pp1.Properties.Add("ID", pi1);
context.Policies.Set<IPropertySetterPolicy>(pp1, typeof(TestObject), "TO1");

PropertySetterInfo pi2 = new PropertySetterInfo("ID", new ValueParameter<string>("ID2"));
PropertySetterPolicy pp2 = new PropertySetterPolicy();
pp2.Properties.Add("ID", pi2);
context.Policies.Set<IPropertySetterPolicy>(pp2, typeof(TestObject), "TO2");

TestObject obj1 = (TestObject)context.HeadOfChain.BuildUp(context, typeof(TestObject),null, "TO1");
TestObject obj2 = (TestObject)context.HeadOfChain.BuildUp(context, typeof(TestObject),null, "TO2");

Console.WriteLine(obj1.ID);
Console.WriteLine(obj2.ID);
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public class TestObject
{
public string _id;

public string ID
{
get
{
return _id;
}
set
{
_id = value;
}
}
}
}

在这个例子中，我们建立了两个PropertySetterPolicy对象，分别以ID2、ID2为id加到了context.Policies中，当CreationStrategy建立对象时，它是以下面的程序代码来取得对应的Policy对象。

private object BuildUpNewObject(IBuilderContext context, Type typeToBuild, object existing, string idToBuild)
{
ICreationPolicy policy = context.Policies.Get<ICreationPolicy>(typeToBuild, idToBuild);
..................
}

这段程序代码告诉我们一个重点，ObjectBuidler是以『类型/id』来做类型识别动作，也就是说TestObject+”ID1”、TestObject+”ID2”被ObjectBuilder视为两个不同的对象建立动作，你可以分别为其设定专属的Policy对象，也可以于调用BuildUp方法时，指定不同的id来建立同类型，但不同id的对象。另一个会使用『类型/id』来做识别的是DependencyResolutionLocatorKey对象，我们之前常使用它来完成Injection动作，而SingletonStrategy、DependencyParameter也都是运用它来完成所需完成的工作，其构造函数如下所示。

public DependencyResolutionLocatorKey(Type type, string id)

这意味着，当我们使用SingletonStrategy时，可以利用『类型/id』来建立两个同类型但不同id的Singleton对象，如程序32所示。

程序32

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_SingletonTwoTest
{
class Program
{
static void Main(string[] args)
{
MyBuilderContext context = new MyBuilderContext();
context.InnerChain.Add(new SingletonStrategy());
context.InnerChain.Add(new CreationStrategy());
context.Policies.Set<ISingletonPolicy>(new SingletonPolicy(true), typeof(TestObject), "ID1");
context.Policies.Set<ISingletonPolicy>(new SingletonPolicy(true), typeof(TestObject), "ID2");
context.Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());

TestObject obj1 = (TestObject)context.HeadOfChain.BuildUp(context, typeof(TestObject), null, "ID1");
TestObject obj2 = (TestObject)context.HeadOfChain.BuildUp(context, typeof(TestObject), null, "ID2");
if (obj1 == obj2)
Console.WriteLine("Singleton");
Console.Read();
}
}

internal class MyBuilderContext : BuilderContext
{
public IReadWriteLocator InnerLocator;
public BuilderStrategyChain InnerChain = new BuilderStrategyChain();
public PolicyList InnerPolicies = new PolicyList();
public LifetimeContainer lifetimeContainer = new LifetimeContainer();

public MyBuilderContext()
: this(new Locator())
{
}

public MyBuilderContext(IReadWriteLocator locator)
{
InnerLocator = locator;
SetLocator(InnerLocator);
StrategyChain = InnerChain;
SetPolicies(InnerPolicies);

if (!Locator.Contains(typeof(ILifetimeContainer)))
Locator.Add(typeof(ILifetimeContainer), lifetimeContainer);
}
}

public class TestObject
{
}
}

这个例子将TestObject+”ID1”、TestObject+”ID2”设定为两个不同的Singleton对象，所以当首次建立并指定id时，所建立出来的两个对象是相异的，也就是说，可以利用『类型/id』来建出两个Singleton系统。

5-5、StrategyList

在本文一开始的范例中，我们使用Builder对象来建立对象，它使用了一个StrategyList对象来处理Strategy串行，这个对象提供了两个重要的方法，一是MakeStrategyChain，它会将StrategyList中的Strategy输出成BuilderStrategyChain对象，这是一个实现了IBuilderStrategyChain接口的对象，也是IBuilderContext所要求的Strategy串行对象。第二个方法是MakeReverseStrategyChain，它会将内含的Strategys反相排序后输出成BuilderStrategyChain对象，这个动作是为了准备TearDown时所需的Strategy串行，还记得前面提过，TearDown的Strategy顺序应该与建立时完全相反，这样才能让对象与其相关的子对象适当的释放。

5-6、TStageEnum

StrategyList是一个泛型对象，它接受一个Enum类型，会依照Enum中所定义的元素来建立Strategy串行或是反相排序，要了解这个设计的原意，我们得先看看ObjectBuilder中所预定义，用来指定给StrategyList的列举。

public enum BuilderStage
{
PreCreation,
Creation,
Initialization,
PostInitialization
}

读者可以查觉，这与我们先前将Strategy分成四种类型的方式相呼应，StrategyList会依据PreCreation、Creation、Initialization、PostInitialization的顺序来产生BuilderStrategyChain对象，这样就不会因为错置Strategy的顺序，导致程序不正常(例如，先加入CreationStrategy再加入TypeMappingStrategy时，TypeMappingStrategy将无法运作)。Builder对象充份展示了BuilderStage与StrategyList的运用方式。

public Builder(IBuilderConfigurator<BuilderStage> configurator)
{
Strategies.AddNew<TypeMappingStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<SingletonStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<ConstructorReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<PropertyReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<MethodReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<CreationStrategy>(BuilderStage.Creation);
Strategies.AddNew<PropertySetterStrategy>(BuilderStage.Initialization);
Strategies.AddNew<MethodExecutionStrategy>(BuilderStage.Initialization);
Strategies.AddNew<BuilderAwareStrategy>(BuilderStage.PostInitialization);
Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());
if (configurator != nulll)
configurator.ApplyConfiguration(this);
}

只要传入的BuilderStage是正确的，不管TypeMappingStrategy是加在CreationStrategy前面或后面，皆可正常运作。不过同一类型的Strategy，但有顺序需求的情况下，仍然要小心调整顺序，程序32示范了运用BuilderStage所带来的优点。

程序32

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_StrategyListTest
{
class Program
{
static void Main(string[] args)
{
MyBuilder builder = new MyBuilder();
ITypeMappingPolicy policy = new TypeMappingPolicy(typeof(TestObject), null);
builder.Policies.Set<ITypeMappingPolicy>(policy, typeof(ITestInterface), null);
ITestInterface obj1 = builder.BuildUp<ITestInterface>(new Locator(), null, null);
Console.Read();
}
}

public class MyBuilder : BuilderBase<BuilderStage>
{
public MyBuilder()
: this(null)
{
}

public MyBuilder(IBuilderConfigurator<BuilderStage> configurator)
{
Strategies.AddNew<CreationStrategy>(BuilderStage.Creation);
Strategies.AddNew<TypeMappingStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<SingletonStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<ConstructorReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<PropertyReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<MethodReflectionStrategy>(BuilderStage.PreCreation);
Strategies.AddNew<PropertySetterStrategy>(BuilderStage.Initialization);
Strategies.AddNew<MethodExecutionStrategy>(BuilderStage.Initialization);
Strategies.AddNew<BuilderAwareStrategy>(BuilderStage.PostInitialization);

Policies.SetDefault<ICreationPolicy>(new DefaultCreationPolicy());

if (configurator != null)
configurator.ApplyConfiguration(this);
}
}

public interface ITestInterface
{
}

public class TestObject : ITestInterface
{
}
}

5-6、PolicyList

BuilderContext所定义的Policies对象类型为PolicyList，PolicyList对象以Dictionary对象来储存设计者所加入的Policy对象，其中用来作为键值的BuilderPolicyKey类别构造函数如下。

public BuilderPolicyKey(Type policyType, Type typePolicyAppliesTo, string idPolicyAppliesTo)

第一个参数为policyType，也就是ICrationPolicy、ITypeMappingPolicy等之类，第二个参数是对应的类型，第三个参数则是id。设计者可以调用PolicyList.Set方法来加入一个Policy至内部的存储器中，该方法会依据传入的参数建立BuilderPolicyKey做为键值，然后将Policy加到Dictionary中，如下所示。

public void Set(Type policyInterface, IBuilderPolicy policy, Type typePolicyAppliesTo,
string idPolicyAppliesTo)
{
BuilderPolicyKey key = new BuilderPolicyKey(policyInterface, typePolicyAppliesTo, idPolicyAppliesTo);
lock (lockObject)
{
policies[key] = policy;
}
}

另一个泛型类型的Set方法也可以达到同样的效果。

public void Set<TPolicyInterface>(TPolicyInterface policy, Type typePolicyAppliesTo,
string idPolicyAppliesTo) where TPolicyInterface : IBuilderPolicy
{
Set(typeof(TPolicyInterface), policy, typePolicyAppliesTo, idPolicyAppliesTo);
}

设计者可以透过PolicyList.Get方法来取得对应的Policy对象，该方法如下所示。

public TPolicyInterface Get<TPolicyInterface>(Type typePolicyAppliesTo, string idPolicyAppliesTo)
where TPolicyInterface : IBuilderPolicy
{
return (TPolicyInterface)Get(typeof(TPolicyInterface), typePolicyAppliesTo, idPolicyAppliesTo);
}

public IBuilderPolicy Get(Type policyInterface, Type typePolicyAppliesTo, string idPolicyAppliesTo)
{
BuilderPolicyKey key = new BuilderPolicyKey(policyInterface,
typePolicyAppliesTo, idPolicyAppliesTo);
lock (lockObject)
{
IBuilderPolicy policy;

if (policies.TryGetValue(key, out policy))
return policy;
BuilderPolicyKey defaultKey = new BuilderPolicyKey(policyInterface, null, null);
if (policies.TryGetValue(defaultKey, out policy))
return policy;
return null;
}
}

SetDefault则可以用一个Policy来提供给所有类型使用，Get方法在找不到对应『类型/id』对应的Policy时，就会以该Policy回传。

六、Locator

ObjectBuilder利用Locator对象来实现Service Locator，也利用Locator来进行Dependency Injection，在ObjectBuilder的架构上，Locator有两种类型，一是Readonly Locator，顾名思义，这类Locator只允许读取、不允许新增。二是ReadWriteLocator，它是允许新增、读取类的Locator。我们可以从Visual Studio 2005的Class Diagram来观察ObjectBuilder中的Locator阶层架构。

图7

6-1、Readonly Locator

ObjectBuidler定义了一个IReadableLocator接口，所有的Locator都必须直接或间接实现此接口，内建实现此接口的类别是ReadableLocator，它是一个抽象类。真正完成实现可用的是ReadOnlyLocator，这个Locator只允许读取，不允许新增。

6-2、ReadWrite Locator

ObjectBuilder中支持读与写的Locator是ReadWriterLocator，与ReadOnlyLocator一样，它也是一个抽象类，真正完成实现的是Locator类别。附带一提，虽然Locator定义了蛮清楚的阶层，但是BuilderContext只支持实现IReadWriterLocator接口的Locator。

6-3、WeakRefDictionary and Locator

Locator类别是我们一直都在使用的Locator，它是一个继承自ReadWriterLocator的类别，值得一提的是，它使用一个WeakRefDictionary来储存设计者所放入Locator的对象，WeakRefDictionary内部对于每个元素都会以WeakReference封装，这意味着，Locator中的元素并无法保证一直都存在，因为CLR会在内存拮据时，先行释放WeakRefernce所封装的对象，这点读者必须谨记。

七、Extending ObjectBuilder

ObjectBuilder除了支持三种Dependency Injection模式、Service Locator之外，最大的魅力应该来自于具高度延展性的架构，设计者可以透过撰写Strategy、Policy、Locator等类别来参与对象的建立动作，本章以两个范例来证明这点，一是EventSetterStrategy，它提供Event Injection功能，二是PoolStrategy，提供Pool模式的对象建立。

7-1、EventSetterStrategy

ObjectBuidler提供了Constructor Injection、Interface Injection(Method Ijection)、Setter Injection(Property Injection)三种Injection模式，虽然ObjectBuilder只提供了Propety式的Setter Injection，不过我们可以藉助于ObjectBuilder高度的延展性架构，让ObjectBuidler也能支持Event Injection。

IEventSetterInfo

Event Injection与Property Injection同属Setter Injection模式，两者运作的模式也极为相似，ObjectBuilder在Property Injection部份是由ProperySeterInfo、PropertySetterPolicy及PropertySetterStrategy三个类别所构筑而成，我们可以依循这个既定架构，实现Event Injection功能。首要必须定义一个IEventSetterInfo接口，这相对于IPropertySetterInfo接口之于Property Injection。

程序33

public interface IEventSetterInfo
{
object GetValue(IBuilderContext context, Type type, string id, EventInfo propInfo);
EventInfo SelectEvent(IBuilderContext context, Type type, string id);
}

IEventSetterInfo接口定义了两个方法，SelectEvent方法是用来取得欲Injection事件的EventInfo对象，EventSetterStrategy会调用此方法来取得欲Injection事件的EventInfo对象，然后透过EventInfo.AddHandler来进行注入动作，这个注入动作所使用的值是透过调用IEventSetterInfo.GetValue方法来取得，此接口的实现程序代码如34。

程序34

public sealed class EventSetterInfo : IEventSetterInfo
{
private string _name = null;
private IParameter _value = null;

#region IEventSetterInfo Members

public object GetValue(IBuilderContext context, Type type, string id, EventInfo propInfo)
{
return _value.GetValue(context);
}

public EventInfo SelectEvent(IBuilderContext context, Type type, string id)
{
return type.GetEvent(_name);
}

#endregion

public EventSetterInfo(string name, IParameter value)
{
_name = name;
_value = value;
}
}

IEventSetterPolicy

前面提过，Strategy是与类型无关的设计，因此需要Policy的协助，我们所设计的EventSetterStrategy也是一样，Event Injection必须具备针对不同『类型/id』进行Event Injection的能力，所以必须设计一个IEventSetterPolicy接口，该接口必须直接或间接继承自IBuilderPolicy接口，这是ObjectBuilder对于Policy的规范。

程序35

public interface IEventSetterPolicy : IBuilderPolicy
{
Dictionary<string, IEventSetterInfo> Events { get;}
}

针对同一『类型/id』对象可能需要注入一个以上的事件，此接口定义了一个Dictionary<string,IEventSetterInfo>对象，让设计者可以指定一个以上的Event Injection动作，36是此接口的实现。

程序36

public sealed class EventSetterPolicy : IEventSetterPolicy
{
private Dictionary<string, IEventSetterInfo> _events = new Dictionary<string, IEventSetterInfo>();

#region IEventPolicy Members
public Dictionary<string, IEventSetterInfo> Events
{
get
{
return _events;
}
}
#endregion
}

EventSetterStrategy

完成了基础类别的设计与实现后，剩下的就是Strategy，也就是EventSetterStrategy的设计与实现了，设计上，EventSetterStrategy只有一个任务，就是于BuildUp方法被调用时，透过『类型/id』经由context.Locator取得对应的IEventSetterPolicy对象，再透过它取得欲进行注入动作的IEventSetterInfo对象，接着调用IEventSetterInfo.SelectEvent方法取得EventInfo对象，最后调用IEventSetterInfo.GetValue取得欲注入的Event Handler对象，然后调用EventInfo.AddHandler方法完成注入动作。

程序37

public class EventSetterStrategy : BuilderStrategy
{
public override object BuildUp(IBuilderContext context, Type typeToBuild,
object existing, string idToBuild)
{
if (existing != null)
InjectEvents(context, existing, idToBuild);

return base.BuildUp(context, typeToBuild, existing, idToBuild);
}

private void InjectEvents(IBuilderContext context, object obj, stringg id)
{
if (obj == null)
return;

Type type = obj.GetType();
IEventSetterPolicy policy = context.Policies.Get<IEventSetterPolicy>(type, id);

if (policy == null)
return;

foreach (IEventSetterInfo eventSetterInfo in policy.Events.Values)
{
EventInfo eventInfo = eventSetterInfo.SelectEvent(context, type, id);

if (eventInfo != null)
{
if (TraceEnabled(context))
TraceBuildUp(context, type, id, "Event Setter", eventInfo.Name);

eventInfo.AddEventHandler(obj,
eventSetterInfo.GetValue(context, type, id, eventInfo) as Delegate);
} }
}
}

Testing

EventSetterStrategy的使用方式与PropertySetterStrategy相似，如38所示。

程序38

using System;
using System.ComponentModel;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace EventSetterTest
{
class Program
{
static void Main(string[] args)
{
Builder builder = new Builder();
builder.Strategies.AddNew<EventSetterStrategy>(BuilderStage.Initialization);
IEventSetterPolicy policy = new EventSetterPolicy();

EventHandler handler = new EventHandler(CallHandler);
policy.Events.Add("Call", new EventSetterInfo("Call",
new ValueParameter(typeof(EventHandler), handler)));
builder.Policies.Set<IEventSetterPolicy>(policy, typeof(TestObject), null);

TestObject obj = builder.BuildUp<TestObject>(new Locator(), null, null);
obj.RaiseCall();
Console.ReadLine();
}

static void CallHandler(object sender, EventArgs args)
{
Console.WriteLine("Called");
}
}

public class TestObject
{
private EventHandlerList _events = new EventHandlerList();
private static object _onCall = new object();

public event EventHandler Call
{
add
{
_events.AddHandler(_onCall, value);
}
remove
{
_events.RemoveHandler(_onCall, value);
}
}

protected virtual void OnCall(EventArgs args)
{
EventHandler handler = (EventHandler)_events[_onCall];
if (handler != null)
handler(this, args);
}

public void RaiseCall()
{
OnCall(EventArgs.Empty);
}
}
}

图8是此程序的运行结果。

图8

7-2、PoolStrategy

GoF的书中，提出了三种对象管理Pattern，一是Singleton，意味着对象一旦建立后，就存放于某个存储器中，之后所有要求对象的建立动作，都将会获得同样的对象实体，在ObjectBuilder中实现这个Pattern的就是SingletonStrategy。第二个Pattern是SingleCall模式，意味所有的对象建立动作都会获得一个新的对象实体，跟new、create等语言所定义的对象建立模式相同，在Service模式中，SingleCall也意味着Service对象会在要求到达时建立，结束后就立即的释放，这两个模式都可以用ObjectBuilder轻易的实现。第三种Pattern就是Pool，也就是说在特定存储器中维持一定数量的对象实体，当要求对象建立动作时，系统会遍寻存储器中的对象，如果有对象标示为未使用状态，那么系统就回传该对象，并将该对象标示为使用中，本节将实现一个PoolStrategy，让ObjectBuilder可以具备Pool的能力。

PoolFactory

Pool Pattern的核心就是一个可以于存储器中管理对象的能力，此处使用笔者书中所设计的PoolFactory类别来完成这个目的。

程序39

using System;
using System.Collections;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace Orphean.WinFormHelper.Framework.Factorys
{
///<summary>
/// a interface to be implement by Object Factory,
/// DAL use object factory to speed object constructing.
///</summary>
public interface IObjectFactory
{
///<summary>
/// acquire a object.
///</summary>
///<param name="type">object Type</param>
///<returns>object</returns>
object AcquireObject(Type type);
///<summary>
/// release a object.
///</summary>
///<param name="obj">a object to releasing</param>
void ReleaseObject(object obj);
}

public sealed class PoolObjectFactory : IObjectFactory, IDisposable
{
class PoolData
{
public bool InUse = falsee;
public object obj;
}

private IList _storage;
private int _max = 100;
private bool _limit = false;
private IBuilderContext _context = null;

public PoolObjectFactory(IBuilderContext context, int max, bool limit, IList storage)
: this(context)
{
_max = max;
_limit = limit;
_storage = storage;
}

public PoolObjectFactory(IBuilderContext context)
{
_context = context;
}

private PoolData GetPoolData(object obj)
{
lock (_storage.SyncRoot)
{
for (int i = 0; i < _storage.Count; i++)
{
PoolData p = (PoolData)_storage[i];
if (p.obj == obj)
return p;
}
}
return null;
}

private object GetObject(Type type)
{
lock (_storage.SyncRoot)
{
if (_storage.Count > 0)
{
if (((PoolData)_storage[0]).obj.GetType() != type)
throw new Exception(
string.Format("the Pool Factory only for Type :{0}", _storage[0].GetType().Name));
}

for (int i = 0; i < _storage.Count; i++)
{
PoolData p = (PoolData)_storage[i];
if (!p.InUse)
{
p.InUse = true;
return p.obj;
}
}

if (_storage.Count > _max && _limit)
throw new Exception("max limit is arrived.");

object obj = _context.HeadOfChain.BuildUp(_context, type, null, null);
PoolData p1 = new PoolData();
p1.InUse = true;
p1.obj = obj;
_storage.Add(p1);
return obj;
}
}

private void PutObject(object obj)
{
PoolData p = GetPoolData(obj);
if (p != null)
p.InUse = false;
}

#region IObjectFactory Members

public object AcquireObject(Type type)
{
return GetObject(type);
}

public void ReleaseObject(object obj)
{
if (_storage.Count > _max)
{
if (obj is IDisposable)
((IDisposable)obj).Dispose();
PoolData p = GetPoolData(obj);
lock (_storage.SyncRoot)
_storage.Remove(p);
return;
}
PutObject(obj);
}

#endregion

#region IDisposable Members

public void Dispose()
{
lock (_storage.SyncRoot)
{
for (int i = 0; i < _storage.Count; i++)
{
PoolData p = (PoolData)_storage[i];
if (p.obj is IDisposable)
((IDisposable)p.obj).Dispose();
}
}
}

#endregion
}
}

本文的重点在于ObjectBuilder的应用与延伸，所以此处就不在赘述PoolFactory的实现细节。

IPoolPolicy

PoolStrategy在架构上与SingletonStrategy类似，我们必须设计一个IPoolPolicy接口，该接口的定义如程序40。

程序40

public interface IPoolPolicy : IBuilderPolicy
{
bool IsPool { get;}
}

此接口只定义了一个Pool属性，用来告诉PoolStrategy那个『类型/id』是需要Pool，那个又是不需要的，虽然设计者可以针对要Pool的『类型/id』来指定IPoolPolicy，如果有特定对象不需要Pool动作，那就不指定IPoolPocy即可，但是我们无法排除一种情况，那就是系统里大多数对象都需要Pool，仅有特定的对象不需要Pool，此时要特别对一个个对象设定IPoolPolicy的话，会相当的繁琐。此时设计者可以以SetDefault来加入IPoolPolicy对象，将所有对象标示为可Pool，再针对不需要Pool的对象来指定IPoolPolicy。程序41是实现此接口的程序代码列表。

程序41

public class PoolPolicy : IPoolPolicy
{
private bool _isPool = false;

#region IPoolPolicy Members

public bool IsPool
{
get
{
return _isPool;
}
}

#endregion

public PoolPolicy(bool isPool)
{
_isPool = isPool;
}
}

PoolStrategy

PoolStrategy必须在BuildUp方法运用PoolFactory来取得要求的对象，在设计上，我们会为每个『类型/id』建立独立的PoolFactory对象，这意味着每个『类型/id』的对象数量是独立管理的。

程序42

using System;
using System.Collections;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;
using Orphean.WinFormHelper.Framework.Factorys;

namespace OB_PoolStrategy
{
public class PoolStrategy : BuilderStrategy
{
private WeakRefDictionary<object, object> _factoryMap =
new WeakRefDictionary<object, object>();
private bool _poolObjectCreating = false;

public override object BuildUp(IBuilderContext context, Type typeToBuild,
object existing, string idToBuild)
{
if (!_poolObjectCreating)
{
IPoolPolicy policy = context.Policies.Get<IPoolPolicy>(typeToBuild, idToBuild);
if (policy != null && policy.IsPool)
{
PoolLocatorKey key = new PoolLocatorKey(typeToBuild, idToBuild);
PoolObjectFactory factory = null;
if (context.Locator.Contains(key))
{
factory = context.Locator.Get<PoolObjectFactory>(key);
lock (this)
{
_poolObjectCreating = true;
try
{
existing = factory.AcquireObject(typeToBuild);
}
finally
{
_poolObjectCreating = false;
}
}
}
else
{
factory = new PoolObjectFactory(context, 15, false, new ArrayList());
_poolObjectCreating = true;
try
{
existing = factory.AcquireObject(typeToBuild);
}
finally
{
_poolObjectCreating = false;
}

context.Locator.Add(key, factory);
}
if (!_factoryMap.ContainsKey(existing))
_factoryMap.Add(existing, factory);
}
}
return base.BuildUp(context, typeToBuild, existing, idToBuild);
}

public override object TearDown(IBuilderContext context, object item)
{
if (_factoryMap.ContainsKey(item))
{
PoolObjectFactory factory = _factoryMap[item] as PoolObjectFactory;
if (factory != null)
factory.ReleaseObject(item);
_factoryMap.Remove(item);
}
return base.TearDown(context, item);
}
}

public sealed class PoolLocatorKey
{
private Type type;
private string id;

public PoolLocatorKey()
: this(null, null)
{
}

public PoolLocatorKey(Type type, string id)
{
this.type = type;
this.id = id;
}

public string ID
{
get { return id; }
}

public Type Type
{
get { return type; }
}

public override bool Equals(object obj)
{
PoolLocatorKey other = obj as PoolLocatorKey;

if (other == null)
return false;

return (Equals(type, other.type) && Equals(id, other.id));
}

public override int GetHashCode()
{
int hashForType = type == null ? 0 : type.GetHashCode();
int hashForID = id == null ? 0 : id.GetHashCode();
return hashForType ^ hashForID;
}
}
}

在BuildUp方法被调用时，PoolStrategy会透过context.Policies取得『类型/id』对应的IPoolPolicy对象，判断此次建立动作是否使用Pool，是的话就以『类型/id』至Locator中取出PoolFactory，如果Locator已经有该PoolFactory时，就直接调用PoolFactory.AcquireObject方法来取得对象实体，如果Locator中无对应的PoolFactory时，就建立一个并放入Locator中。在这个建立流程中有几个重点，第一！我们将PoolFactory储存在Locator中，因此需要一个类似DependencyResolutionLocatorKey的对象，用来做为由Locator取出PoolFactory的键值，这个对象必须覆载Equal、GetHashCode两个方法，因为Locator会调用这两个方法来比对键值，这个对象就是PoolLocatorKey。第二！PoolFactory在存储器中没有可使用对象时，会调用BuilderContext.HeadChain.BuildUp方法来建立该对象，这会引发重进入的问题，BuilderContext.HeadChain.BuildUp方法将会再次触发PoolStrategy的BuildUp，而这里又会再次调用BuilderContext.HeadChain.BuildUp，造成重入的问题，所以此处利用一个旗标：poolObjectCreating来解决这个问题。第三！PoolStrategy必须在TearDown方法被调用时，调用PoolFactory.ReleaseObject来将该对象归还，此时会遭遇到一个问题，因为TearDown方法只会传入对象实体，没有id的信息，这使得PoolStrategy无法于此处取得对应的PoolFactory对象，为了解决此问题，PoolStrategy宣告了一个_factoryMap对象，它是一个WeakRefDictionary<object, object>类别对象，在对象实体于BuildUp方法被建立后，PoolStrategy会将object/PoolFactory成对放入_factoryMap中，这样就能于TearDown时以对象实体至_factoryMap中取出对应的PoolFactory对象了。

Testing

PoolStrategy的使用方式与SingletonStrategy类似，程序43是应用的程序代码列表。

程序43

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.Practices.ObjectBuilder;

namespace OB_PoolStrategy
{
class Program
{
static void Main(string[] args)
{
Builder builder = new Builder();
builder.Strategies.AddNew<PoolStrategy>(BuilderStage.PreCreation);

IPoolPolicy policy = new PoolPolicy(true);
builder.Policies.Set<IPoolPolicy>(policy, typeof(TestObject), null);

Locator locator = new Locator();
TestObject obj1 = builder.BuildUp<TestObject>(locator, null, null);
TestObject obj2 = builder.BuildUp<TestObject>(locator, null, null);
builder.TearDown<TestObject>(locator, obj1);
builder.TearDown<TestObject>(locator, obj2);
TestObject obj3 = builder.BuildUp<TestObject>(locator, null, null);
if (obj3 == obj1 || obj3 == obj2)
Console.WriteLine("Pooled");
Console.ReadLine();

}
}

public class TestObject
{
}
}

图9是执行结果。

图9